Download Figure 1: Casseopeia A (Cambridge 1

Figure 1:
Casseopeia A (Cambridge 1-mile)
Figure 2:
Casseopeia A (Cambridge 1-mile)
Figure 3:
Cygnus A (Cambridge 1-mile)
Figure 4:
Cygnus A (Cambridge 1-mile)
Figure 5:
The Very Large Array (VLA)
Figure 6:
Casseopeia A (VLA)
Figure 7:
Cygnus A (VLA)
Figure 8:
Cygnus A (VLA)
Figure 9:
Cygnus A (VLA)
Figure 10:
The Very Long Baseline Array (VLBA)
Figure 11:
How a radio telescope functions
Very Long Baseline Interferometry
Richard Porcas
Max-Planck-Institut fuer Radioastronomie, Bonn
1
Contents
• Introduction
• Principles and Practice of VLBI
• High angular resolution of long baselines
• The “geophysics” of long baselines
• Interferometer connection over long distances
• Recorded signals and off-line correlation
• The “geopolitics” of long baselines
• Some useful AIPS tasks for VLBI analysis
2
Introduction
• Very Long Baseline Interferometry: VLBI
• Use longest possible baselines on the Earth
• Use arrays of ∼10 antennas: EVN and VLBA
• Use baseline from Earth to a radio telescope on an orbiting
satellite: VSOP mission.
• Higher angular resolution imaging for finer structural detail
• Higher precision astrometry
• Traditionally realized without real-time signal links between
antennas
• Local autonomous telescope control, executing a pre-distributed
observing schedule to synchronise telescopes in the array
• Independent local oscillator signals
• Local sampling of voltage signals and recording on tapes (and
disks)
• Later off-line replay of sampled voltage signals for multiplication at a correlator; JIVE, Bonn, VLBA-Socorro
• Modern realization of real-time VLBI
3
Principles and Practice of VLBI
• Basic principles are exactly the same as for any other radio
interferometry.
• Realization of long baselines involves practical differences
It is instructive to group the reasons why these arise under a
number of headings:
• High angular resolution of long baselines
• The “geophysics” of long baselines
• Interferometer connection over long distances
• Recorded signals and off-line correlation
• The “geopolitics” of long baselines !
4
High angular resolution of long baselines
• Longer baselines magnify the effect of small angular quantities
• Increased phase difference between 2 points in the sky produces better angular resolution of target source !
Need small pixels ! AIPS IMAGR, CELLSIZE=0.0004
• But also on amplitude calibrators - they may be heavily resolved
Generally cannot rely on point source amplitude calibrators
• Calibration by radiometry (SEFD) of individual antennas
We calibrate the “noise” level and hence the correlation in Jy
AIPS tasks ANTAB APCAL CLCAL
• May need to correct for bias in the digital representation of the
voltage if the sampler levels are not correct. This correction
is done at the EVN correlator at JIVE, but not at the VLBA
correlator.
AIPS tasks FITLD (option DIGICOR=) and ACCOR
• Larger holes in the u,v plane
“Missing flux” if interferometer array does not have sufficiently short baselines
• Characteristic brightness temperature lower limit for detection given by shortest baseline in the array.
Source must be sufficiently compact (i.e. have sufficiently high
brightness temperature) to be detectable
• This limit is much higher for VLBI because the shortest baselines are usually longer !
Tb > 108 - 1012; synchrotron emission, maser emission
• Increased path difference produced by small angular errors
Need more accurate angular geometry to ensure path difference is less than 1/b
• More accurate source positions
1 arcsec with 5 km baseline; 2.4 cm (144o at 5 GHz)
1 arcsec with 5000 km baseline; 24 m (80 nanoseconds; 12 MHz)
Smaller delay beam; decreased field of view
• Earth Orientation Parameters (UT1-UTC, position of the pole)
GR deflection due to the Sun is important over whole sky
• Need finer frequency sampling at the correlator (0.5 MHz)
Can be achieved with multi-lag correlation
• Increase rate-of-change of phase errors
Residual fringe rate, f
Cannot average data longer than a time 1/f
• Smaller fringe-rate beam; decreased field of view
• Need finer time sampling at the correlator (1 s)
May need shorter time intervals in AIPS CL table
• Finer time and frequency sampling of the visibility function
increases the volume of data !
• As with other interferometers we must calibrate phase errors.
• Phase self-calibration leads to a detection threshold i.e. must
have sufficient signal-to-noise ratio in the solution interval
• Need to average data in frequency and time to achieve this
• Need to determine the unknown residual delay and residual
fringe-rate to permit this averaging
• Fringe Fitting procedure is used to determine the residual
fringe rate and delay
• Forms averages of the data in time and frequency with a range
of trial fringe rates and delays, and determines which values
maximise the signal.
• Requires enough signal-to-noise to recognise the signal so sets
a minimum correlated flux density threshold for detection
• AIPS task FRING does a “global” search for all baselines simultaneously, making use of the fact that the residual delay
and fringe rate on a baseline may be expressed as the difference between antenna-based delay and rate residuals
• FRING also incorporates a phase self-calibration solution with
a default point-source model
• For an N-antenna array, FRING determines (N-1) antennabased phase, fringe-rate and delay solutions with respect to a
reference antenna whose residuals are assigned to zero.
• Note that the act of averaging the visibility function, using
the residual delays and fringe rates determined by FRING,
decreases the field of view of the averaged data.
5
The “geophysics” of long baselines
• The Earth is round
• It is surrounded with a layer of neutral and ionized gas
• It is not entirely solid
• Source has different elevations at different antennas
Short time of mutual visibility for low declination sources
Galactic Centre source not visible on transatlantic baselines
• Program SCHED for planning observing sequence
• For a VLBI array there may be much more data for short
baselines than for long baselines
• Parallactic angles are different for azimuth/elevation mounted
antennas at different sites.
• Need to use feeds which respond to a circular polarization
mode to get correlation
• The parallactic angle difference for the 2 antennas of a baseline
introduces an additional phase term which is not corrected by
the correlator, and is of opposite sign for RHC and LHC
• It may be necessary to correct for this when making
polarization VLBI observations, or when
averaging LHC and RHC correlations together, or when doing
phase-reference observations if the target and reference source
have different parallactic angles.
AIPS task CLCOR, option PANG
• Non-equal tropospheric paths
• For very short baselines (< 5 km) we assume that the tropospheric paths of the signals to the 2 antennas have the same
physical length and that the temporal fluctuations along them
are the same. The increased path lengths, and temporal phase
fluctuations, cancel.
• For longer baselines (> 10 km) we assume the tropospheric
paths have the same physical length but that temporal phase
fluctuations are independent along the 2 paths. The increased
path lengths cancel but the phase fluctations do not, and appear in the measured visibility function.
They must by removed (be phase-referencing or phase selfcalibration)
• For VLBI the physical paths through the troposphere have
very different lengths because of the different elevations of
the antennas.
• The increased path lengths do not cancel but produce a relative delay, τ , which is a function of the zenith delays, z and
the elevations e for the 2 antennas (A,B):
τ = zA/cos(eA) − zB /cos(eB )
In order to equalize the paths for correlation, a model of the
tropospheric delay is needed.
• Phase-reference observations
Target and phase-reference source observations are alternated,
and the phase measured on the reference is used to calibrate
the tropospheric phase fluctuations on the target.
• Note that τtarget and τref erence are not equal and it is therefore
essential to have a model for the tropospheric delay path.
• Non-equal ionospheric paths
More significant at low frequencies
Varies with time of day, latitude, Solar cycle......
“Zenith delay” increase can be very different for different antennas
• It is unpredictable so no correction is made for this by VLBI
correlators.
• For phase-referenced observations it may be necessary to correct for the ionospheric path increases after correlation, using
an ionospheric model derived from GPS data.
AIPS task TECOR
• Telescopes may be on different tectonic plates
Telescope baselines can change by ∼ cm per year
• Telescopes at different longitudes experience different displacements due to solid Earth tides (∼ 10 cm).
6
Interferometer connection over long distances
• Could control telescopes centrally in real time over telephone/internet
lines (but we don’t..)
• Cannot easily use cables, waveguides or microwave links to
bring the wideband (e.g. 128 MHz) signals together.
• Cannot easily distribute common LO signal
(but may use satellite link for SKA)
• Use independent LOs derived from Hydrogen maser frequency
standard
Phase stability of 1 part in 1014
• Sampling of signal voltages controlled by accurate local clock.
• Encode time signals to identify each sampled bit
• Accurate monitoring of observatory clocks using GPS (∼100 nanoseconds)
• eVLBI: Bring bit-streams together via internet, or satellite,
or dedicated optical fibres
• Real-time correlation, with suitably large buffer to accommodate large astronomical and terrestrial delay between bitstreams.
Corresponding bits must be matched and multiplied
• Need to take account of local clock and clock-rate differences
using fringe fitting
• eVLBI is being developed at present
7
Recorded signals and off-line correlation
• VLBI traditionally performed by recording the bit-streams
and the encoded clock information on magnetic tape (and now
computer disks)
• Replace cables, optical fibres, internet with airplanes, ships
and trucks
• Assemble all the recorded bit-streams at the correlator, some
weeks later
• Synchronised replay of the recorded bit-streams under computer control
• The correlator need not know they are not in real time !
• Tape recorded radio interferometry was the only possibility in
the late 1960’s when VLBI was first performed
• The main difficulty is to write, record and replay a sufficiently
high bit-rate
• Bit-rate = twice the bandwidth
• Initially done using computer tape (∼100 kHz bandwidth)
• 1974: NRAO Mk2 system; 2 MHz (4 Mbps)
Using professional TV video recorders (later VCRs)
• Further translation of IF signal to baseband (0 - 2 MHz) using
baseband converter (also video-converter) before sampling the
voltages.
• 1979: Haystack Mk3 system; 28 Mk2s in parallel
28 separate 2 MHz channels (112 Mbps)
14 baseband converters (upper and lower sidebands)
longitudinal recording on 1” magnetic tape at 3m/s with 28
parallel tracks !
• 1980’s VLBA and
8 separate 8 MHz
4 separate 8 MHz
8 separate 8 MHz
etc....
Mk4 recording systems
channels (128 Mb/s for 1 bit per sample)
channels (128 Mb/s for 2 bits per sample)
channels (256 Mb/s for 2 bits per sample)
• Can have 4 channels RHC and 4 channels LHC for polarization
VLBI
• Can have 4 channels 2.3 GHz and 4 channels 8.4 GHz
• 2004: Mk5 disk-based recording
Increases possible bit-rate to 1024 Mbps
• The main effect from using recorded interferometry comes
from the (traditional) use of multiple channels of relatively
narrow bandwidth (in AIPS “multiple IFs”)
• Different basebands have different instrumental phase errors
• The instrumental phase differences between basebands are
generally fairly constant with time during an experiment
• Need to calibrate these phase offsets in order to average the
“IFs” to get better signal-to-noise ratio (e.g. for FRING or
CALIB)
• VLBA has phase-cal system; a series of standard tones is inserted in the RF signal band and the phases are measured in
each baseband.
• AIPS tasks PCLOD and PCCOR are used to apply these
phase corrections
• In the absence of a phase-cal signal, a strong source can be
used to determine the relative phase between basebands.
Known as manual phase-cal
Use FRING to determine instrumental phase in each “IF”
separately
• The use of “off-line” correlation instead of realtime correlation
has little effect for most observations.
• Correlation can be repeated if replay is poor or if mistakes are
made !
• Multiple-pass correlation can be performed in certain special
cases:
If the number of telescopes exceeds the number of playback
units (> 16 for JIVE; > 20 for VLBA)
If 2 separate field-centres in the primary telescope beams are
to be correlated
• The data on the tape delivered from the correlator may not
be in strict time-baseline, tb order
Use AIPS task MSORT to order it if desired
8
The “geopolitics” of long baselines
• Long baselines tend to cross international borders !
• Leads naturally to international collaboration for performing
VLBI observations.
Example: the European VLBI Network (EVN)
• Use of recorded VLBI makes it easy to use a variety of different
radio telescopes, built for other purposes.
Telescopes in such an array can have a large range of sizes,
sensitivities, beamwidths, etc
• Different baselines in the array can have very different sensitivities
Need to consider carefully how to weight the data from different baselines when performing mapping and self-calibration
In AIPS, consider “calibrate the weights” (DOCALIB=2)
• Organization and operation of an international array such as
the EVN has to involve many different national funding agencies (unlike the US VLBA).
Telescopes available for the EVN depends on observing frequency.
• Some observing options:
• EVN (9+ telescopes at 1.7 GHz and 5 GHz, 8+ at 6 GHz)
• VLBA (10 telescopes at a range of frequencies)
• Global array of VLBA + EVN (16+ telescopes)
• Global 3mm VLBI array (13 telescopes)
• The High Sensitivity Array (HSA):
VLBA + GBT + VLA + ARECIBO + EFFELSBERG
9
Some useful AIPS tasks for VLBI analysis
• FITLD to read in data (DIGICOR option for VLBA correlator)
• MSORT if data not in TB order
• FXPOL is sometimes needed to fix wrongly-labelled polarization channels from the VLBA correlator !
• ACCOR determines amplitude corrections from sampling level
errors for each antenna and writes an SN table (for VLBA correlator)
• CLCOR with option PANG to make parallactic angle correction in a CL table if necessary
• ANTAB reads external file of amplitude calibration radiometry and produces TY (system temperature) and GC (gain
curve) tables.
• APCAL extracts amplitude calibration data from TY and GC
tables and writes an SN table.
• PCLOD reads external file of phase-cal data and writes a PC
table
• PCCOR extracts phase-cal information from PC table and
writes an SN table
• CLCAL interpolates SN table values into a CL (calibration)
table
• FRING determines (antenna-based) residual phase, delay and
fringe-rate by self-calibration and writes an SN table. Can be
used to determine “IF” instrumental phase offsets.
• CALIB for self-calibration of antenna phases and amplitudes
• SPLIT to apply accumulated calibration (maybe calibrate the
weights) and to average frequencies (and possibly “IFs”)